In a multi-part assembly, the order of welding decides how much freedom each joint has when it is made. A joint welded while its surroundings are still flexible can pull the assembly into a new shape. A joint welded after the assembly has become stiff has to resolve its shrinkage inside a rigid frame, so it stores stress instead of moving.
That gives a simple principle: weld the joints that are allowed to move first, and the joints that must hold tolerance last. It is not a rule about which joint is more important. It is a rule about which joint has the least freedom to move when it is welded.
Most fabrication problems that get blamed on a difficult joint come from this principle being inverted. The stiff joint is welded first because it is easiest to reach, and every later weld then fights it.
A tube stub welded onto an open shell is relatively free. The shell can flex, the stub can move, and the shrinkage resolves as a small change of shape in a part that has nothing to hold. Weld the same stub after the end caps and the internal baffle are in, and the shell cannot flex, so the stub pulls and distorts the cap.
This is why internal structure goes in later than intuition suggests. Baffles, stiffeners and gussets are restraint. They should be added once the outer geometry is established, not before, unless the design needs them as a fixture to hold a shape during welding.
The exception is a part that is too floppy to weld without internal support. In that case the stiffener goes in early as tooling and its own weld is sequenced to minimise its effect. The distinction is whether the stiffener is structural or temporary.
A sub-assembly welded on its own bench can move freely. A sub-assembly welded inside a larger structure cannot. Building the small unit first, checking it, and then fitting it into the parent gives the shop a chance to correct the unit before its geometry is locked.
It also splits the heat. A large assembly welded in one continuous campaign accumulates heat across the whole structure, and the last joints run far hotter than the first. Breaking the work into stages allows cooling between them, which reduces the accumulated movement.
The cost is handling and re-fixturing, and that is a real cost. On low-volume work the trade still usually favours sub-assembly, because a correction on a small unit is cheap and a correction on a finished assembly is not.
Within each stage, the heat should be distributed so the part does not develop a hot side and a cold side. Welding all the joints on one face and then moving to the other face is the common mistake. The first face shrinks while the second is cold, the part bows, and the second face then welds a bow into place.
The alternative is to alternate. Weld a joint on one side, move to the opposite side, and return. On symmetrical assemblies this is straightforward. On asymmetric ones it means deciding which side has more deposit and starting there, so the larger pull is balanced by the smaller one later.
Stitch welds complicate the picture. Where a part is stitched at 80 x 60 mm along a seam, the individual stitches are short and the heat input per stitch is modest, but the sequence of stitches still sets the direction of movement. Running stitches from one end to the other accumulates; alternating between ends balances.
Some sequences are decided by the torch rather than the metallurgy. A joint that can only be reached before a neighbouring part is fitted has to be welded early, whatever the shrinkage argument says. Pretending otherwise produces a drawing that cannot be built.
The way to handle this is to identify the access-locked joints first and treat them as fixed points, then sequence everything else around them. On a manifold or a multi-branch assembly this is usually the branch-to-collector joints, since the collector closes access as it goes on.
It also helps to leave one access route open until the end. A part that is fully closed has no way to reach an internal joint for repair, and a repair attempt through a small opening usually makes the damage worse.
Where an assembly includes a machined face, a gasket surface or a bolt pattern, the safest sequence welds everything that will move before that feature is finished. Welding after machining means the machining tolerance is consumed by weld movement.
If the design requires machining after welding, then the sequence must be finished and stable, and the part should be allowed to cool and settle before it goes on the machine. Rushing a warm part onto a mill produces a face that is flat at the time and not flat afterwards.
The same reasoning applies to holes. Drill after welding where possible, or drill undersize and finish later. A hole drilled before a nearby weld is a hole that will move.
A sequence that lives in one welder's head is not a process. It needs to be on the job card as a numbered list of joints, with the direction of travel for each, and it needs to be short enough that somebody will actually follow it.
Photographs of the assembly with the joints numbered are the most practical format, because they remove the ambiguity of describing a joint in words. On repeat work the photograph becomes part of the work instruction.
The sequence should also record what was measured between stages. That is what turns a one-off success into a repeatable allowance, and it is the only way to know whether a change in the sequence improved or damaged the result.
For multi-part weldments, the drawing should indicate which features are critical and which joints are structural, because those two facts determine the sequence.
Multi-part welded assemblies and brackets are listed under other parts, and tube-based structures under exhaust pipe. Titanium fabrication work sits on titanium fabrication components, with welded details on TIG titanium welding parts. The drawing, sample and approval route is described on customization, and production capacity on about us.
